ZeroShotOpt: Towards Zero-Shot Pretrained Models for Efficient Black-Box Optimization

2025-10-03

AI Summary

Bayesian optimization (BO) methods for global optimization of expensive, derivative-free black-box functions suffer from poor generalization and reliance on manual hyperparameter tuning. Method: We propose the first pre-trained policy model for continuous black-box optimization, trained offline via reinforcement learning on over one million synthetic Gaussian process functions and diverse BO variant trajectories. This large-scale pre-training enables zero-shot transferable optimization policies without fine-tuning. Contribution/Results: The method achieves state-of-the-art sample efficiency across multiple benchmark suites—outperforming or matching leading BO and global optimization baselines—while demonstrating superior generalization, robustness, and scalability. Crucially, it introduces the pre-training paradigm to black-box optimization, establishing a new foundation for scalable, adaptive optimization. To foster reproducibility and further research, we publicly release our code, datasets, and pre-trained models.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Optimization
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Global optimization of expensive, derivative-free black-box functions requires extreme sample efficiency. While Bayesian optimization (BO) is the current state-of-the-art, its performance hinges on surrogate and acquisition function hyper-parameters that are often hand-tuned and fail to generalize across problem landscapes. We present ZeroShotOpt, a general-purpose, pretrained model for continuous black-box optimization tasks ranging from 2D to 20D. Our approach leverages offline reinforcement learning on large-scale optimization trajectories collected from 12 BO variants. To scale pretraining, we generate millions of synthetic Gaussian process-based functions with diverse landscapes, enabling the model to learn transferable optimization policies. As a result, ZeroShotOpt achieves robust zero-shot generalization on a wide array of unseen benchmarks, matching or surpassing the sample efficiency of leading global optimizers, including BO, while also offering a reusable foundation for future extensions and improvements. Our open-source code, dataset, and model are available at: https://github.com/jamisonmeindl/zeroshotopt

Problem

• Optimizing expensive black-box functions without derivatives efficiently
• Overcoming Bayesian optimization's reliance on hand-tuned hyperparameters
• Achieving robust zero-shot generalization across diverse problem landscapes

Innovation

• Pretrained model using offline reinforcement learning
• Generates synthetic Gaussian process-based functions
• Achieves zero-shot generalization on unseen benchmarks